Ranked-choice voting is a form of voting in which electors place candidates in an order of preference rather than selecting only one. The expression describes a ballot format, but its meaning varies by jurisdiction. In contemporary American usage, it usually denotes instant-runoff voting, a single-winner system that successively eliminates the lowest-ranked candidates and transfers their ballots to remaining choices. It can also include the single transferable vote, a related method for electing several representatives. More broadly, ranked ballots can be counted by several other rules, which need not produce the same winner.
The distinction between a ballot and a counting rule is fundamental. Ranking candidates records information about relative preferences; the electoral rule determines how that information becomes an outcome. Single-winner ranked-choice voting generally seeks a winner with support extending beyond a first-choice plurality, whereas multiwinner transferable voting can provide proportional representation. These methods have different consequences for electoral competition, representation, campaign strategy, and administration.
Preferential systems have been used in national elections in Australia and Ireland for more than a century, alongside applications in local elections, organizational elections, and American state and municipal contests. Their analysis combines questions of democratic representation with practical issues such as ballot design, transfer procedures, exhausted ballots, and the mathematical limits of collective choice.
Terminology and the information contained in a ranking
A ranked ballot usually asks a voter to assign the number 1 to a first choice, 2 to a second choice, and so forth, or to mark candidates in numbered columns. A ranking such as indicates that the voter prefers candidate A to B and B to C. It does not establish how strongly the voter holds either preference. Someone who considers A excellent and B barely acceptable can submit the same order as someone who regards them as almost indistinguishable. Ranked voting therefore records ordinal preferences rather than numerical measures of satisfaction.
The expression “preferential voting” is common in Australia and other Commonwealth settings. “Alternative vote” usually means single-winner instant-runoff voting, especially in British discussions. “Ranked-choice voting” has become prominent in the United States, often as a replacement for “instant runoff.” None of these terms should be treated as a universal specification. An election law must identify the number of seats, permissible rankings, counting procedure, and treatment of incomplete or ambiguous ballots.
Ranking can be compulsory, optional, or partially compulsory. Under full preferential voting, an elector normally must rank every candidate for a valid vote, subject to applicable saving provisions. Under optional preferential voting, a valid ballot may name only a first choice. Other systems require a specified minimum number of preferences or allow voters to rank only a fixed number of candidates. These distinctions affect both ballot completion and the number of votes that remain transferable late in a count. They also separate compulsory ranking from compulsory attendance or participation: the latter concerns whether a person must take part in the election, not how preferences must be expressed.
Ranked ballots are distinct from approval voting, in which a voter can support several candidates without ordering them, and from score voting, in which candidates receive numerical ratings. They also differ from the conventional two-round system, which requires a later election if no candidate meets the initial winning condition. An instant-runoff ballot supplies advance instructions for successive transfers, but it cannot reproduce every feature of a separate runoff. Between two actual elections, voters may change their minds, candidates may make endorsements, campaign information may change, and turnout may differ.
Some ranked methods use every recorded preference in comparisons or calculations. Others consult a lower preference only after a higher-ranked candidate has been eliminated or elected. Consequently, claims about “ranked voting” need to distinguish properties of ranking itself from properties of a particular counting algorithm.
Single-winner counting and ballot exhaustion
Under the usual instant-runoff rule, each ballot initially counts for its highest-ranked eligible candidate. If a candidate has more than half of the votes in the relevant continuing total, that candidate wins. Otherwise, the candidate with the fewest votes is eliminated. Each ballot assigned to that candidate transfers to its next-ranked candidate still in contention. Elimination and transfer continue until a winner is determined under the jurisdiction’s rules. Some procedures stop when a majority is reached; others complete a final comparison between the last two candidates.
Consider 100 valid ballots divided into three groups: 42 rank A first, 33 rank B first and C second, and 25 rank C first and B second. No candidate initially has a majority. C is eliminated, and all 25 of C’s ballots transfer to B. B therefore wins with 58 votes against A’s 42. A would have won under plurality voting, because plurality requires only more first-choice votes than any opponent. The example illustrates how transfers can change the outcome; it does not imply that the initial leader normally loses in actual ranked-choice elections.
An exhausted ballot is one that contains no further usable preference for a continuing candidate. Suppose the 25 C-first ballots in the example instead contain no additional rankings. After C’s elimination, only 75 ballots remain active: A has 42 and B has 33. A then has 56 percent of continuing votes, although only 42 percent of the original valid ballots. Thus, a “majority winner” under optional ranking may mean a majority of the ballots still participating in the decisive round, not a majority of everyone who cast a valid first preference. This denominator is important when relating the result to majority rule.
Exhaustion is not necessarily evidence of an error. A voter may intentionally decline to support any remaining candidate. It can also result from limits on the number of permitted rankings or from an ambiguous mark that prevents further interpretation. A skipped rank, repeated candidate, or equal ranking can have different effects under different statutes. Many jurisdictions preserve an unambiguous first preference even if a later part of the ballot cannot be used.
Tie-breaking and elimination procedures are also substantive rules. Ties may be resolved through previous-round totals, a statutory ordering rule, or drawing lots. Some systems permit simultaneous elimination of several candidates when their combined votes are too few to overtake the next candidate. Such shortcuts require safeguards to preserve the result of the prescribed sequential count. Official reporting commonly presents round-by-round totals so that the path from first preferences to the final result can be reconstructed.
Multiwinner ranked voting and proportional representation
The single transferable vote, usually abbreviated STV, uses rankings to elect several candidates in the same district. Unlike single-winner instant-runoff voting, it distributes representation among groups of voters rather than identifying just one winner. Votes can transfer both from eliminated candidates and from candidates who have more support than they need for election. This prevents all the votes of a large group from being consumed in electing one representative when some of that voting power can help elect another.
A common election threshold is the Droop quota. For valid votes and seats, its integer form is
In a three-seat election with 100 valid votes, the quota is 26. Four candidates cannot each obtain 26 votes from that total, so no more than three candidates can independently reach the quota. A candidate who reaches the quota is elected. Any surplus above the quota is transferred according to later preferences using the jurisdiction’s transfer procedure. If no further candidate can be elected, the lowest candidate is eliminated and the ballots transfer. Counting continues until all seats are filled; remaining candidates can sometimes be declared elected without reaching the quota when their number equals the vacancies.
Surplus distribution is technically more complex than ordinary elimination transfers. Under a fractional method, transferable ballots can pass onward at a reduced value. In a simplified case where a candidate has 40 whole-value votes, the quota is 26, and every ballot can transfer, the surplus is 14 and each ballot can transfer at a value of . Actual rules differ over which ballots enter a surplus calculation, how previously transferred values are handled, and how rounding occurs. Some historical and current systems select physical ballot papers for transfer rather than transferring an identical fraction of every relevant paper. Different STV implementations are therefore not interchangeable.
The proportional effect depends strongly on district magnitude—the number of representatives elected together. With three seats, the Droop quota is slightly above one-quarter of the valid vote; with five seats, it is slightly above one-sixth. Larger districts generally permit smaller groups to secure representation, although candidate preferences and transfers also matter. STV is not identical to nationwide party-list proportional representation. It provides proportionality within districts, and national outcomes also depend on district boundaries, seat allocation, and geographical patterns of support.
STV gives voters choices among candidates of the same party as well as between parties, and it accommodates independent candidates without requiring a party list. Conversely, electing several candidates through repeated single-winner instant runoffs is not necessarily proportional. If the same majority can choose every winner, minority groups may remain unrepresented. The proportional character of STV comes from its quota and transfer structure, not merely from allowing rankings or filling several seats.
Intellectual origins and institutional development
Modern ranked voting emerged through several distinct lines of electoral reform. In 1770, Jean-Charles de Borda presented the ranked scoring method now called the Borda count to the French Academy of Sciences. Under its familiar form, candidates receive points according to their positions on each ballot. In 1785, the Marquis de Condorcet examined the comparison of alternatives by majority preference and the possibility that collective preferences form a cycle. These approaches use rankings, but neither is the same as the later elimination-and-transfer systems.
The nineteenth-century development of transferable voting was closely associated with proposals for proportional representation. Carl Christopher Georg Andræ devised a transferable system used in Denmark in the 1850s. In Britain, Thomas Hare published proposals in 1857 and developed them further in his 1859 treatise on representation. John Stuart Mill promoted Hare’s approach in Considerations on Representative Government in 1861. These reformers sought to allow substantial bodies of electors to obtain representation even when they were dispersed rather than concentrated in a territorial majority.
Thomas Hare’s original proposals differed from many modern district-based implementations. They envisaged much larger electoral arrangements and raised questions about how voters would identify candidates and how votes would be transferred. Henry Richmond Droop subsequently developed the quota associated with his name. Other reformers refined counting procedures and sought systems manageable with paper ballots. The history of STV is therefore one of successive institutional and computational adaptations rather than the invention of a single unchanging procedure.
William Robert Ware proposed a single-winner elimination-and-transfer method in 1871. This adaptation dispensed with proportional seat allocation and instead used successive eliminations to choose one candidate. Its later development into the alternative vote was particularly important in Australia. Tasmania introduced transferable voting for elections in Hobart and Launceston in 1896, and the Hare–Clark system was extended statewide for its House of Assembly election in 1909. Australia adopted preferential voting for federal House of Representatives elections in 1918.
Adoption has not always been permanent or uniform. Several American cities introduced STV during the early twentieth century, often as part of municipal reform, but many subsequently repealed it. Cambridge, Massachusetts, which first used proportional ranked voting for municipal elections in 1941, became a durable American example. In other countries, ranked methods became embedded in constitutional or statutory arrangements. This history includes both campaigns for broader representation and campaigns for decisive single-member outcomes; those aims explain why preferential voting developed into more than one institutional family.
National and comparative applications
Australia illustrates the coexistence of different preferential systems within one national political structure. The federal House of Representatives uses full preferential voting in single-member divisions. Electors ordinarily rank all candidates, and counting identifies a winner through first preferences and transfers. The federal Senate instead uses proportional STV in statewide and territory-wide electorates. Since the 2016 Senate reforms, ballot instructions have asked electors to rank at least six groups above the line or at least twelve individual candidates below the line, with statutory provisions preserving some ballots that do not meet those instructions.
These chambers demonstrate why ballot format alone does not determine representation. In the House, each division returns one member, and votes cast elsewhere cannot help elect that division’s representative. In ordinary half-Senate elections, each state elects six senators, creating opportunities for several parties to gain seats. When all twelve state Senate seats are contested at a double dissolution election, the quota is lower. State and territory electoral systems add further variation: Tasmania and the Australian Capital Territory use Hare–Clark STV for their lower or sole legislative chambers, while other jurisdictions employ different combinations of preferential and proportional rules.
Ireland uses STV for elections to Dáil Éireann, its lower parliamentary chamber, in constituencies returning three, four, or five members. Voters rank individual candidates, and transfers can pass between candidates of the same party or across party lines. Ireland’s president is elected by a single-winner preferential method. Proposals to replace Dáil STV with single-member plurality voting were rejected in referendums in 1959 and 1968. Malta also uses STV for its national parliament, although constitutional seat-adjustment mechanisms mean that its final party seat totals cannot be understood from constituency counts alone.
Northern Ireland uses STV for Assembly and local council elections, and Scottish local councils have used STV since 2007. Elsewhere, preferential rules can take forms unlike standard instant runoff. Papua New Guinea uses limited preferential voting, introduced for national general elections in 2007, in which voters are required to indicate three preferences. Sri Lanka’s presidential system allows voters to express additional preferences, but its procedure, if no candidate receives an initial majority, compares the top two candidates using eligible later preferences rather than conducting the usual series of lowest-candidate eliminations.
The United Kingdom held a referendum in 2011 on replacing plurality voting with the alternative vote for House of Commons elections. The proposal was rejected. This and other examples show that ranked voting can be adopted for one institution while remaining absent from others. Comparisons must therefore specify the office, election date, district size, and governing rules rather than treating entire countries as having one homogeneous voting system.
Adoption and legal design in the United States
American use of ranked-choice voting expanded during the early twenty-first century, especially for single-winner municipal elections. San Francisco voters approved a reform in 2002, and the city first used it in 2004. Minneapolis began using ranked-choice voting in 2009. New York City approved its reform in 2019 and first used it in citywide election cycles in 2021 for municipal primary and special elections. Its system permits up to five candidate rankings. It does not replace the city’s ordinary partisan general elections with ranked-choice voting.
Maine’s experience illustrates the interaction between electoral reform and a state constitution. Voters approved ranked-choice voting in 2016. In a 2017 advisory opinion, the state’s highest court concluded that applying it to specified state general elections conflicted with constitutional provisions prescribing election by plurality. The subsequent legal framework allowed ranked-choice voting in relevant primaries and federal elections while leaving those constitutional constraints in place. In the 2018 election for Maine’s second congressional district, Jared Golden defeated Bruce Poliquin after transfers, despite trailing him in the initial first-choice count.
Alaska adopted a different package through a 2020 ballot initiative and first used it in statewide elections in 2022. Its system combines a nonpartisan top-four primary election with a ranked-choice general election. The primary advances four candidates regardless of party, and the general election determines the winner through rankings. In the August 2022 special election for Alaska’s United States House seat, Mary Peltola won after the elimination of Nick Begich and the distribution of available later preferences between Peltola and Sarah Palin. The primary structure and the ranked count are analytically separate reforms, even though Alaska enacted them together.
Legal controversies commonly concern the interpretation of majority requirements, state authority over municipal elections, ballot validity, and the relationship between successive rounds and equal voting rights. The fact that a ballot can transfer does not mean its voter receives several simultaneous votes: in an ordinary instant-runoff count, each active ballot contributes to only one continuing candidate at a time. Nevertheless, legal conclusions depend on the applicable constitutional and statutory text. A procedure permitted under one state’s law need not be available to a municipality in another state.
American reforms also differ in scope. Some apply only to partisan nomination, some to nonpartisan municipal offices, and others to federal general elections. Certain jurisdictions use ranked ballots for military or overseas voters as a contingency if a later runoff occurs, without using ranked-choice tabulation for the entire electorate. Adoption, repeal, and state restrictions make the national picture change over time. A meaningful account therefore distinguishes the approval of a proposal, its effective date, the elections actually conducted under it, and any later legal changes.
Mathematical properties and competing definitions of fairness
Voting rules embody different standards of collective choice. The Condorcet criterion requires electing a candidate who would defeat every rival in a separate head-to-head majority comparison, whenever such a candidate exists. Instant-runoff voting does not always satisfy this criterion. A broadly acceptable candidate can have relatively few first preferences and be eliminated before later preferences reveal that the candidate would defeat each remaining rival. By contrast, a Condorcet method explicitly uses pairwise comparisons, although it needs additional rules when no candidate defeats every opponent.
The 2009 mayoral election in Burlington, Vermont, became a prominent example. Bob Kiss won the instant-runoff count, while rankings on the ballots indicated that Andy Montroll would have defeated each of the other candidates in a head-to-head comparison. Montroll was eliminated before the final round. Burlington subsequently repealed that mayoral system in 2010. The case demonstrates a possible divergence between two definitions of majority support: prevailing among the final continuing candidates and defeating every candidate in separate pairwise comparisons.
Instant-runoff voting also fails the monotonicity criterion in some preference configurations. Raising a candidate on some ballots, without otherwise changing those ballots’ relative ordering, can alter which opponent is eliminated first and thereby cause that candidate to lose. Conversely, lowering a candidate can sometimes help the candidate win. These possibilities arise because elimination order determines which preferences become operative. Their mathematical existence does not establish their frequency in public elections; estimating frequency requires complete ballot data and a specified definition of the relevant change.
Another important property is later-no-harm. Under standard single-winner instant runoff, expressing a later preference cannot cause a candidate ranked earlier on that same ballot to lose. Lower preferences are not consulted while the earlier choice remains active. This provides a reason to distinguish ranking a fallback from abandoning a first choice. Nevertheless, instant runoff is not fully strategy-proof: in some circumstances a voter can obtain a preferred outcome by reporting an insincere ranking. Satisfying later-no-harm does not eliminate every possible incentive for tactical behavior.
Arrow’s impossibility theorem places such disputes within a broader theory of collective preference: under its assumptions, no rule converting unrestricted individual rankings into a complete social ordering satisfies all the specified conditions. The Gibbard–Satterthwaite theorem establishes a related strategic limitation for deterministic, non-dictatorial single-winner rules with at least three possible outcomes and unrestricted preferences. These theorems do not make electoral design arbitrary. They show why systems must be compared through explicit priorities—such as proportionality, pairwise majority support, simplicity, or resistance to particular strategies—rather than an expectation that one method can satisfy every desirable condition.
Voter behavior, campaigning, and representation
Ranked-choice voting changes the choices available to voters, but institutional effects are not automatic. Under single-winner plurality, a voter may abandon a preferred minor candidate to support a more competitive alternative. Instant runoff allows a ballot to begin with the preferred candidate and transfer if that candidate is eliminated. This can reduce some forms of vote splitting. It does not eliminate every “spoiler” effect, because introducing or removing a candidate can change the elimination order and the eventual winner even when that candidate does not win.
Candidates may have incentives to seek second and subsequent preferences from supporters of rivals. Campaigns can emphasize shared positions, exchange recommendations about rankings, or appeal beyond a core constituency. In a transferable-vote system, an endorsement can specify a sequence rather than only a first choice. Whether these incentives reduce negative campaigning depends on electoral competition, party organization, media coverage, and voter behavior. A candidate can still benefit from attacking a close competitor, particularly if the immediate objective is to survive an early elimination.
The relationship to party systems also depends on the counting rule. Single-member instant runoff remains a winner-take-all district system and does not, by itself, allocate legislative seats proportionally to national vote shares. Australia’s House of Representatives demonstrates that preferential voting can coexist with competition centered on two major political blocs. Multiwinner STV creates different opportunities because a group need not command a district majority to elect a representative. Even there, district magnitude, geography, nomination decisions, and transfers shape the number of viable parties.
Voter participation and ballot completion involve separate questions. A ranked ballot may require more candidate information and more marking than a single-choice ballot. Its actual difficulty depends on the number of candidates, layout, instructions, and whether rankings are optional. A ballot containing only one preference may be entirely valid under optional rules, so single-ranking rates cannot simply be treated as error rates. Similarly, exhaustion rates should distinguish deliberate truncation, ranking limits, and invalid or ambiguous later preferences.
Research on turnout, demographic disparities, campaign tone, and candidate diversity faces substantial problems of causal inference. Cities adopting reform may differ from other cities before implementation; elections held at different times attract different electorates; and closely associated changes may alter nomination or election schedules. Useful studies separate these factors and distinguish a finding in one jurisdiction from a universal effect. Candidate diversity, in particular, depends on recruitment, campaign resources, party support, and the distribution of voters as well as the counting rule. The ability to rank candidates creates opportunities for different electoral strategies, but it does not guarantee a particular social composition of officeholders.
Election administration, transparency, and auditing
Administering a ranked election begins with ballot design. Numbered columns must make clear which direction represents first, second, and later choices. Instructions need to distinguish ranking several candidates from giving the same rank to several candidates. Paper size, candidate order, language access, and accommodations for voters with disabilities can affect usability. Electronic ballot-marking devices may prevent some invalid combinations, but the voter-verifiable record still needs to express the intended ranking clearly. An accessible interface and an accurate tabulation system are separate administrative requirements.
Counting ranked ballots requires retaining more information than a simple precinct total for each candidate. First-choice totals alone do not reveal how ballots will transfer. Election officials may count manually, use software to process records of individual ballots, or combine both approaches. STV adds fractional values, surplus-transfer rules, and potentially lengthy calculations. The governing specification must establish precision and rounding, the order of operations, treatment of ties, and the circumstances in which candidates are declared elected. Software is implementing those legal rules, not supplying a substitute for them.
Results can be slower to finalize when later-arriving eligible ballots could alter the elimination sequence. This does not necessarily mean that the transfer computation itself is slow. Scanning, signature verification where required, adjudication of ambiguous marks, and deadlines for receiving ballots may account for much of the elapsed time. Publishing preliminary first preferences is also different from publishing a completed ranked count. Clear reporting identifies how many ballots remain outstanding and avoids treating an early transfer sequence as irreversible when the underlying ballot set is incomplete.
Transparency commonly involves releasing round-by-round totals, explanations of exhausted votes, and, where permitted, anonymized cast-vote records. Such records allow independent analysts to reproduce a tabulation, but they raise privacy questions. A long and unusual ranking can function as a distinctive pattern, especially if records are linked to small precincts or detailed ballot-order information. Election authorities must balance reproducibility with ballot secrecy and applicable disclosure law. Reproducing published electronic records establishes consistency of the computation; it does not alone establish that those records accurately reflect the paper ballots.
Post-election verification therefore needs to examine both ballot interpretation and the outcome. Risk-limiting audits use statistical procedures to limit the chance of leaving an incorrect reported outcome uncorrected. Ranked elections can require checking several assertions about elimination order or candidate comparisons rather than merely confirming one final margin. Multiwinner STV presents additional difficulties because surplus transfers and earlier elections affect later stages. Secure chain of custody, preserved paper records, documented adjudication, and a legally specified recount procedure remain essential regardless of how the rankings are tabulated.